Tracking instruction-tuned LLM openness. Paper: Liesenfeld, Andreas, Alianda Lopez, and Mark Dingemanse. 2023. “Opening up ChatGPT: Tracking Openness, Transparency, and Accountability in Instruction-Tuned Text Generators.” In Proceedings of the 5th International Conference on Conversational User Interfaces. doi:10.1145/3571884.3604316.
https://mistral.ai/news/mixtral-of-experts/
Very interesting from a performance standpoint: https://twitter.com/lmsysorg/status/1734680611393073289 :eyes:
Huggingface: https://huggingface.co/blog/mixtral
Note re: openness: https://twitter.com/karpathy/status/1734251375163511203